AI Smart Assistant on AWS: Personalized Shopping Experience
05. august 2025 ┃ 5 minút čítania
Challenge
The retailer faced pressure from a changing market, where success no longer depended solely on the breadth of the product range, but on the ability to respond quickly and accurately to customer needs. Online visitors expected the same service as in a brick-and-mortar store, but without an assistant to advise them on product selection and compatibility, the decision-making process was prolonged and the willingness to complete a purchase declined. There was a need to connect product data, staff knowledge, and the digital channel into a single intelligent interface.
Comprehensive range: extensive product catalog, including furnishings, tiles, and kitchen systems, which complicated online decision-making.
Low online conversion: customers often abandoned their shopping carts due to a lack of personalized assistance.
Risk of losing competitiveness: digital competitors with fast, personalized experiences were gaining market share.
Unused data: weak link between customer interactions and marketing and product development.
Solution
Before deploying AI Smart Assistant, alternatives such as a classic recommendation system and extended full-text search were considered. However, these approaches were unable to respond to the context of customer questions, explain product selection, or adapt to individual preferences. A solution was needed that would combine data accuracy with natural interaction and be able to leverage existing catalog and customer data without interfering with the infrastructure.
Aspecta designed and deployed AI Smart Assistant — a digital shopping advisor available 24/7 that combines conversational AI, semantic search, and integration with e-shop and ERP systems.
Main technologies (AWS-native):
Amazon Bedrock (large context models) – generative AI for conversations and recommendations.
Amazon OpenSearch (Vector Engine) – semantic search and vector queries.
Amazon API Gateway + AWS Lambda / ECS – orchestration and middleware.
Amazon DynamoDB – persistence of sessions, and context.
Amazon S3 – centralized storage of catalogs, knowledge bases, and logs.
CloudTrail, GuardDuty, CloudWatch – security and observability.
Integration: Middleware ensures secure connection to ERP and e-shop without interfering with the client's core systems.
Implementation
The architecture was built in a multi-account AWS environment (dev/test/prod) with an emphasis on scalability and isolation.
Phase:
PoC: model validation and data mapping.
Pilot: deployment for selected product categories.
The Rollout: extension to the entire catalog and integration with payments/inventory.
Operation: monitoring, iterations, and regular updates of knowledge bases.
Duration: approximately 6 months from PoC to production launch (reference project).
Results and benefits
After a successful rollout, the project became key evidence that properly designed AI can bring not only technological innovation but also direct business impact. The assistant connected customer behavior, data, and recommendation mechanisms into a single ecosystem, enabling immediate measurement of the effect and rapid optimization. The following metrics illustrate how these changes translated into real results:
+18% relative increase in online conversions in the first 6 months.
~30% improvement in product search and recommendation success rates.
+12 points improvement in NPS for online interactions.
Higher average basket value thanks to relevant cross-selling.
Generating valuable KYC insights for segmentation and product development.
Lower operating costs thanks to serverless and managed services.
Why the solution worked:
Fast time-to-value thanks to managed AWS services (Bedrock, OpenSearch).
Modular architecture enabled quick updates to the catalog and prompts without downtime.
Secure integrations via middleware minimized the risk to the client's existing systems.
Measurable KPIs linked to business metrics (conversion, NPS, AOV).
"The solution delivered the expected business benefits — fast online customer support and new data for the marketing and product teams."
— Head of Digital Transformation
Conclusion
The project confirmed that the integration of generative AI and AWS data services can truly change the way customers shop online. The solution delivered a rapid return on investment (ROI) thanks to increased conversions, optimized marketing costs, and reduced need for manual support. The combination of scalable cloud architecture and contextual AI created a sustainable foundation for further expansion—from voice assistants to personalized real-time recommendations.
Do you need to transform online shopping in your e-shop into an intuitive, conversion-driven experience? Don't hesitate to contact us.
Subscribe for newsletter
Share article
